A neural architecture, fuzzy ARTMAP (Carpenter et al 1992), is considered here as an alternative to standard feedforward networks for noisy mapping tasks. It is one of a series of architectures based upon adaptive resonance theory or ART (Carpener et al 1991a; 1991b; 1992). Like other ART based systems, fuzzy ARTMAP has advantages over feedforward networks and is especially suited to classification-type problems. Here, it is used to approximate a noisy mapping. Results show that properties which confer useful advantages for classification problems do not necessarily confer similar advantages for noisy mapping problems. One particular feature, match-tracking, is found to cause over-learning of the data. A modified variant is proposed, without match-tracking, which stores probability information in the map field. This information is subsequently used to commute output estimates. The proposed fuzzy ARTMAP variant is found to outperform fuzzy ARTMAP in a mapping task.
展开▼